In the last three months I’ve been re-acquainting myself with basic statistics. It’s a periodic re-learning that usually occurs when I don’t get the statistical result I had wanted or suddenly develop an urge to learn R. This time around, I was struck with a realization: no scientist ever uses the null hypothesis.
I want to use this blog post to make a case for using this very same null hypothesis.
I think the argument boils down to the idea that nothing in science can ever be proven true. That is nature of the scientific method and the statistical analyses we commonly use. The best that we can hope to do is disprove a negative. This, to me, makes a lot of intuitive sense. The nature of science is to never be certain and always questioning results because new techniques and findings are bound to come along that will reshape our understanding. Nothing is definite, which makes disproving a negative all the more practical that trying to come up with a definite.
Personally, I think that there is an psychological advantage to using null hypotheses. If you’re never in search of a positive you are less likely to be disappointed with the inevitable negative. Its a matter of perspective. I believe that if we train ourselves to only try and disprove no effect hypotheses, then we would never try to chase false positives. A real problem considering that our favorite P-values don’t really mean much.
So I ask, what’s in a hypothesis? Statistically, it is required to perform the necessary calculations and make assumptions but I don’t know of any scientist that strictly adheres to these principles. I think that for most of us, it is a fluid idea that shifts and changes with every experiment. We are all trying to form a coherent and logical story based off of our results.
But perhaps we shouldn’t. For the sake of doing good science and preserving our own mental health it might be better to have a definite preconceived null hypotheses to disprove.